They would have to do biocomp testing. Every 510(k) requires an analysis of device compliance to ISO 10993, regardless of whether or not its predicate is preamendment. The premise of the article is alarmist and acts as if FDA requires no testing whatsoever, when the list of standards is as long as my arm: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpcd/cla...
Now one could argue that biocomp testing, which is done on animal models for the most part, is insufficient for any implanted device. That's a fair argument and one worth having. But the linked op-ed is a polemic.